eurekin

20 ms·

271.
by eurekin 3y ago
In case of programming, there could be some mileage with synthetically generated code examples.
272.
by eurekin 3y ago
Oh! That's actually cool. Could use that mixtral on the next pc as well
273.
by eurekin 3y ago
How do you go about that (besides manually editing the whole prompt)?
274.
by eurekin 3y ago
I'm using the chatgpt sometimes 2-3 hours a day. That could easily go 10x
275.
by eurekin 3y ago
I burn through my 10 USD limit in a weekend or two, if I'm using the API. Is that not an issue for you?
276.
by eurekin 3y ago
Confirming I've been using it extensively in the past and now, for same kind of chats, I hit limits much easily (never before), since most of messages contain "// here you put the actual implementation" and I'm
277.
by eurekin 3y ago
Speculating How about a chat interface, where you correct the result and provide more contextual information about those columns? Those chats could be later fed back to the model and ran a DPO optimisation on top
278.
by eurekin 3y ago
At that time, there still will be a "A100/H100/successor replacement specialist" probably necessary. Or any other variant of a data centre hardware admin
279.
by eurekin 3y ago
AI
280.
by eurekin 3y ago
Hi! How's that working for you? I have a similar cluster, but for CEPH specifically and I couldn't bear the hassle (supposedly hassless updates stuck, until I cleaned up some dirs as a root etc.)
281.
by eurekin 3y ago
oh wow! I've cought that system wide find on a few occasions and thought it was related to the specific of my employer. AFAIRC, it was scanning for vulnerable log4j versions
282.
by eurekin 3y ago
Ok, I didn't see that: I saw a person watching gore that turned out to be one of the most helpful and selfless persons I know of. Doesn't fit any definition of a sociopath I looked up, quite contrary
283.
by eurekin 3y ago
I'd love to see training working on arm + mps macs. Hopefully someone knowledeable chimes in, that it's actually possible. I tried running some mpx supported conda/pytorch inference (whisper to be precise) on a workplace issu
284.
by eurekin 3y ago
Did I do that?
285.
by eurekin 3y ago
To provide a more expanded horizon: it was a hobby of a peculiar group of kids in my middle school (and earlier). They teased each other, who can be more cool and unfazed (eat a sandwich for example), while watching it. One of them is a par
286.
by eurekin 3y ago
That's exactly my impression of his image as well
287.
by eurekin 3y ago
They seem to be following the same circulating memo :) Obviously must work, otherwise they would not have done it
288.
by eurekin 3y ago
I read that in the exact same way: things are still more costly than performant so we need a stable injection
289.
by eurekin 3y ago
It really would not work with 14 plus?
290.
by eurekin 3y ago
For the purpose of learning, would you recommend some tutorials, articles or videos that help achieve that? Accuracy aside, this would make a great learning experience! Is it better to look in the PyTorch community, or that's where som
291.
by eurekin 3y ago
Thanks! Would they also help with reading the exact rpm and speed values?
292.
by eurekin 3y ago
I'm suspecting it could be similar to the learned vs. predefined positional embeddings in GPTs. That is, the learned version is a "warped and distorted" version of the exact predefined pattern, and yet somehow it performs a b
293.
by eurekin 3y ago
Just a single anecdotal data point, how good the result can be: https://www.youtube.com/watch?v=Nnpm-rJfFjQ The accuracy is probably quite bad, when looked at individual manipulations, but the overall result is jaw dropping
294.
by eurekin 3y ago
To be fair, I have no idea how I could use modern techniques instead of the Hough Transform. One use case is the speed and RPM car dials recognition. Using the Hough Transform, it's trivial to reliably detect the slope with a high degr
295.
by eurekin 3y ago
I can't say for sure now, but one time I was helping with a similar project for bank agents working "in the field." The core requirement was that data was strictly confidential and could not be processed anywhere but where th
296.
by eurekin 3y ago
Yes, but still valuable. It's a good read and I learned something new, despite doing exactly that project ~10 years ago :)
297.
by eurekin 3y ago
How are you accessing it?
298.
by eurekin 3y ago
Care to share, what are you using it for? I'm curious, because I'm gathering some usecases; so that I could share that internally in the company to provide better education on, what LLMs do and how they work.
299.
by eurekin 3y ago
Theranos resemblance is off the charts
300.
by eurekin 3y ago
About 3., isn't that done specifically in order to not overfit?

More ›